Fast or Slow, Either Head Can Start the Processive Run of Kinesin-2 KIF3AC*

Mammalian KIF3AC contains two distinct motor polypeptides and is best known for its role in organelle transport in neurons. Our recent studies showed that KIF3AC is as processive as conventional kinesin-1, suggesting that their ATPase mechanochemistry may be similar. However, the presence of two different motor polypeptides in KIF3AC implies that there must be a cellular advantage for the KIF3AC heterodimer. The hypothesis tested was whether there is an intrinsic bias within KIF3AC such that either KIF3A or KIF3C initiates the processive run. To pursue these experiments, a mechanistic approach was used to compare the pre-steady-state kinetics of KIF3AC to the kinetics of homodimeric KIF3AA and KIF3CC. The results indicate that microtubule collision at 11.4 μm−1 s−1 coupled with ADP release at 78 s−1 are fast steps for homodimeric KIF3AA. In contrast, KIF3CC exhibits much slower microtubule association at 2.1 μm−1 s−1 and ADP release at 8 s−1. For KIF3AC, microtubule association at 6.6 μm−1 s−1 and ADP release at 51 s−1 are intermediate between the constants for KIF3AA and KIF3CC. These results indicate that either KIF3A or KIF3C can initiate the processive run. Surprisingly, the kinetics of the initial event of microtubule collision followed by ADP release for KIF3AC is not equivalent to 1:1 mixtures of KIF3AA plus KIF3CC homodimers at the same motor concentration. These results reveal that the intermolecular communication within the KIF3AC heterodimer modulates entry into the processive run regardless of whether the run is initiated by the KIF3A or KIF3C motor domain.

for KIF3A when expressed as a heterodimer with KIF3C includes the native sequence of the motor domain, the neck linker, and helix ␣7 (Met 1 -Leu 374 ), followed by the dimerization motif of EB1 (DFYFGKLRNIELICQENEGENDPVLQRIV-DILYATDE) and the C-terminal tobacco etch virus proteasecleavable Strep tag (TTSENLYFQGASNWSHPQFEK). The KIF3C construct for KIF3AC includes the native sequence of the KIF3C motor domain, neck linker, and helix ␣7 (Met 1 -Leu 374 ), followed by the dimerization motif of EB1 (DFYF-GKLRNIELICQENEGENDPVLQRIVDILYATDE) and the tobacco etch virus protease-cleavable His 8 tag (TTSENLYFQ-GASHHHHHHHH). The KIF3AA homodimer contained the same sequence as the construct for the heterodimer, except the purification tag used was the His 8 tag. The plasmid for KIF3CC homodimer expression is the same as for heterodimer expression. The resulting constructs were sequence-verified over the entire ORF insert (30).
KIF3 Protein Expression and Purification-KIF3 motors were expressed in Escherichia coli BL21-CodonPlus (DE3)-RIL cell line (Stratagene, La Jolla, CA) for purification as described in detail previously (30,32). To achieve the purification of KIF3AC heterodimers, sequential affinity columns were used. The HisTrap FF Ni ϩ2 -NTA column (GE Healthcare) selected for KIF3C because of its C-terminal His 8 tag, and the StrepTactin TM column (GE Healthcare) selected for the Strep-tagged KIF3A. The purified KIF3AC was evaluated by analytical gel filtration chromatography (Superose TM 10/300; GE Healthcare) and SDS-PAGE to confirm purification of stable heterodimers with a 1:1 stoichiometry of KIF3A to KIF3C. The predicted molecular weight based on amino acid sequence of KIF3AC is 98,317.
To purify homodimeric KIF3AA or KIF3CC, the supernatant of the cell lysate was loaded onto the HisTrap FF Ni ϩ2 -NTA column (GE Healthcare) to select for the specific homodimer expressed. The purified homodimers of KIF3AA or KIF3CC were analyzed by analytical gel filtration and SDS-PAGE. The predicted molecular weights based on amino acid sequence of KIF3AA and KIF3CC are 97,000 and 99,518, respectively.
The KIF3 proteins were dialyzed against ATPase buffer containing 20 mM Hepes, pH 7.2, with KOH, 5 mM magnesium acetate, 0.1 mM EDTA, 0.1 mM EGTA, 50 mM potassium acetate, 1 mM dithiothreitol, plus 5% sucrose followed by ultracentrifugation to remove aggregates and then frozen as aliquots in liquid N 2 for storage at Ϫ80°C. Before each experiment, KIF3 protein aliquots were thawed and clarified for 10 min at 4°C (Beckman Coulter TLX Optima Ultracentrifuge, TLA-100 rotor, 313,000 ϫ g), and the protein concentration was determined using the Bio-Rad protein assay with IgG as a protein standard. The experiment is designed using the KIF3 dimer concentration but converted to the ATP site concentration (two sites per dimer) to report here.
Experimental Conditions-The experiments were performed at 25°C in ATPase buffer: 20 mM Hepes, pH 7.2, with KOH, 5 mM magnesium acetate, 0.1 mM EDTA, 0.1 mM EGTA, 50 mM potassium acetate, 1 mM dithiothreitol, plus 5% sucrose. The morning of each experiment bovine tubulin was cold-depolymerized, clarified, and polymerized with 1 mM MgGTP at 37°C, followed by MT 2 stabilization with 40 M paclitaxel. The MT concentrations reported represent those of the paclitaxelstabilized tubulin polymer. The reported concentrations of ATP, GTP, and the nucleotide analog mantATP include an equivalent concentration of magnesium acetate. Note that the fluorescent analog mantATP was purchased from Invitrogen as the isomeric mixture. The errors are reported as S.E.
MantATP Binding Kinetics-To determine the second order rate constant of ATP binding, the fluorescent ATP analog, mantATP (Invitrogen), was used. The MT⅐KIF3 complexes were preformed and rapidly mixed with mantATP in the stopped-flow instrument (KinTek SF2003; KinTek Corp., Austin, TX). The change in the relative fluorescence of mantATP was monitored by excitation at 360 nm with detection at ϳ450 nm using a 409-nm long pass filter (Semrock Inc.). Upon mantATP binding to the hydrophobic pocket of the active site, the fluorescence was enhanced, and a double exponential function was fit to each biphasic transient (see Fig. 1). The observed rates of the initial fast exponential phase of each transient were plotted as a function of increasing mantATP concentration. The following equation was fit to the data, where k obs is the observed rate of the initial exponential phase, k ϩ1 represents the second order rate constant for mantATP binding, and k Ϫ1 is the dissociation rate constant obtained from the y intercept. The observed rates of the second exponential phase of each transient were also plotted as a function of mantATP concentration, and a hyperbolic function was fit to these data (see Fig. 1, Scheme 1, and Table 1).
Pulse-Chase Kinetics of ATP Binding-To determine the time course of ATP binding, a chemical quench-flow instrument (RQF-3, KinTek Corp.) was used (32,33). The MT⅐KIF3AC complex (5 M KIF3AC/40 M MTs or 10 M KIF3AC/40 M MTs, syringe concentrations) was rapidly mixed with MgATP plus radiolabeled [␣-32 P]ATP and 300 mM KCl (syringe concentrations) for times ranging from 5 to 200 ms in the instrument. The reaction was subsequently chased with 30 mM nonradiolabeled MgATP plus 150 mM KCl (syringe concentrations) for 0.7 s (Ͼ10 ATP turnovers) and then expelled from the instrument into an Eppendorf tube containing 100 l of 22 N formic acid. The acid terminates the reaction, unfolds the protein, and releases nucleotide from the active site. During the time of chase, stably bound [␣-32 P]ATP was converted to [␣-32 P]ADP⅐P i , whereas any [␣-32 P]ATP unbound or bound loosely to the active site was diluted by the high concentration of nonradiolabeled ATP. The nucleotide products ([␣-32 P]ADP ϩ P i ) were separated from the substrate [␣-32 P]ATP by thin layer chromatography and quantified by using Image Gauge V4 software (FUJIFILM Science Laboratory). The con-centration of [␣-32 P]ADP product was plotted as a function of time. Equation 2 was fit to each transient, where A 0 is the amplitude of the exponential phase, representing the concentration of stably bound ATP that proceeded through ATP hydrolysis during the first ATP turnover, k b is the observed rate of the exponential phase, and t is the time in seconds. The k slow is the rate constant of the linear phase (M⅐s Ϫ1 ) and, when divided by the active site concentration, corresponds to steady-state turnover at the conditions of the experiment. The observed exponential rates of each transient were plotted as a function of MgATP concentration, and Equation 3 was fit to the data, where K 1 represents the equilibrium association constant for collision complex formation (K d ϭ 1/K 1 ), and kЈ ϩ1 is the first order rate constant for the ATP-promoted isomerization. The amplitude of each transient was plotted as a function of increasing MgATP concentration, and the hyperbolic fit to the data provided the maximum amplitude of the exponential phase (see Fig. 2, A-C, Scheme 1, and Table 1).
Acid-quench Kinetics of ATP Hydrolysis-To measure the time course of ATP hydrolysis, the MT⅐KIF3AC complex (10 M KIF3AC and 40 M MTs, syringe concentrations) was rapidly mixed with MgATP plus trace [␣-32 P]ATP and 300 mM KCl (syringe concentrations) for times ranging from 5 to 300 ms in a KinTek chemical quench-flow instrument (32,33). The reaction was terminated with 22 N formic acid and expelled from the instrument. The products [␣-32 P]ADP ϩ P i were separated from [␣-32 P]ATP and quantified, and the concentration of [␣-32 P]ADP was plotted as a function of time. Equation 2 was fit to each transient, yielding the observed rate and amplitude, which were plotted as a function of MgATP concentration. A hyperbolic function was fit to each data set, thereby providing the rate constant of ATP hydrolysis and the maximum amplitude of the initial exponential phase that corresponds to the first ATP turnover (see Fig. 2, D-F, Scheme 1, and Table 1).
ADP Release upon MT Collision-To measure the kinetics of ADP release upon MT collision, the experimental design mimicked the reaction condition at the beginning of the cycle (see Fig. 5, E0 -E1). KIF3 motors were rapidly mixed in the stoppedflow instrument with MTs plus mantATP. At the conditions of the assay, mantATP binding was significantly faster than ADP release from the active site upon MT collision. Therefore, the enhanced fluorescence upon mantATP binding was used as a readout of ADP release. KIF3 at 0.5, 1, or 5 M sites was rapidly mixed with 0. A double exponential function was fit to each transient. The observed rates of the initial exponential phase were plotted as a function of MT concentration, and the hyperbolic fit to the data provided the maximum rate constant of ADP release. At low concentrations of MTs, the observed rate of ADP release is limited by MT collision; therefore, the linear fit to these data provided the second order rate constant of MT association (see Fig. 3, Scheme 1, and Table 1).
Is There an Intrinsic Bias of Which KIF3 Head Initiates the Processive Run?-To test this hypothesis, additional ADP release experiments were performed as described in the previous section. KIF3AC, KIF3AA, KIF3CC, and mixtures of KIF3AA ϩ KIF3CC at 5 M KIF3 site concentration were rapidly mixed with 30 M MTs plus 100 M mantATP (syringe concentrations). The final concentrations were 2.5 M KIF3 site concentration, 15 M MTs, and 50 M mantATP. The mixtures corresponded to 75% KIF3AA ϩ 25% KIF3CC, 50% KIF3AA ϩ 50% KIF3CC, and 25% KIF3AA ϩ 75% KIF3CC. A double exponential fit to each transient provided the rate constants and relative amplitudes of each exponential phase (see Fig. 4).
To further analyze the transients of the KIF3AA and KIF3CC mixtures of homodimers, equivalent transients were generated computationally. Each stopped-flow transient represents a thousand data points during the collection time, with each data point representing the relative fluorescence output for the KIF3 sites. Therefore, the experimental transients can be evaluated using the data from KIF3AA and KIF3CC transients to computationally generate a transient at the appropriate composition of the KIF3AA ϩ KIF3CC experimental curve. For example, to generate a computationally derived transient for 50% KIF3AA ϩ 50% KIF3CC, the data for the time courses of 100% KIF3AA and 100% KIF3CC were each divided by 2 and summed to generate a computationally derived transient. These computationally derived transients were then compared directly to the experimental transients for the mixtures and the transient for KIF3AC (see Fig. 4B).

Results
Scheme 1 is the KIF3AC ATPase pathway used to design the pre-steady-state kinetics experiments, and Fig. 5 presents a stepping model of the KIF3AC interactions with the MT. The rate and equilibrium constants are reported in Table 1.
MantATP Binding Kinetics-To determine the kinetics of ATP binding, two approaches were used. In the first, the fluorescence enhancement of the analog mantATP binding to the KIF3 motors was monitored (Fig. 1). For the second (Fig. 2, A-C), a pulse-chase approach with radiolabeled ATP was used. For the mantATP binding kinetics, the preformed MT⅐KIF3 complex was rapidly mixed in the stopped-flow instrument with increasing concentrations of mantATP. Fig. 1A shows representative transients of mantATP binding to the MT⅐KIF3AC complex, and the biphasic kinetics required a double exponential fit to each transient. The observed rates of the initial exponential phase were plotted as a function of increasing mantATP concentration (Fig. 1B), and the linear fit of the data provided the second order rate constant of mantATP binding at 11 M Ϫ1 s Ϫ1 . The inset in Fig. 1B reveals that the second phase of mantATP binding was also mantATP concentration-dependent, and the hyperbolic fit to the data provided the rate of 22 s Ϫ1 . Because the steady-state k cat for KIF3AC is 21.5 s Ϫ1 , these results are consistent with an alternating site model for KIF3AC as proposed in Fig. 5.
The results in Fig. 1C show that mantATP also binds the MT⅐KIF3AA complex rapidly at 16 M Ϫ1 s Ϫ1 . In contrast, KIF3CC displays a very slow second order rate constant for mantATP binding at 0.68 M Ϫ1 s Ϫ1 . These results are consistent with our previous studies in which the steady-state ATPase kinetics and single molecule velocities for KIF3AA were faster than those for KIF3AC, and KIF3CC was a dramatically slower motor (Table 1 and Ref. 30). The mantATP results also reveal that at this early step in the pathway (Fig. 5, E1-E2), the KIF3A head is very different from the KIF3C head based on the results for homodimeric KIF3AA and KIF3CC.
ATP Binding and ATP Hydrolysis Are Both Fast Events for KIF3AC-The kinetics of ATP binding were also measured by the pulse-chase approach using radiolabeled ATP (Fig. 2, A-C). The MT⅐KIF3AC complex was preformed and rapidly mixed with [␣-32 P]ATP in the chemical quench-flow instrument for times from 5 to 200 ms and subsequently chased with a high concentration of unlabeled MgATP for 0.7 s. During the time of the ATP chase, unbound [␣-32 P]ATP and loosely bound [␣-32 P]ATP were diluted by the high concentration of unlabeled ATP. However, [␣-32 P]ATP stably bound to the active site proceeds through ATP hydrolysis. Therefore, this experimental design quantifies the MT⅐KIF3AC*⅐[␣-32 P]ATP intermediate that proceeds through ATP hydrolysis (Scheme 1). Because of the high processivity of KIF3AC, an additional 300 mM KCl (syringe concentration) was added to the [␣-32 P]ATP syringe, and 150 mM KCl was added to the ATP chase syringe. The additional salt does not affect the first turnover but slows steady-state ATP turnover and therefore enables the initial exponential burst phase to be better resolved.
Each transient displayed two phases: the initial exponential phase corresponding to the first ATP turnover and the following linear phase representing subsequent ATP binding events or steady-state activity ( Fig. 2A). The observed rates of the initial exponential phase of each transient were plotted as a function of ATP concentration. Equation 3 was fit to the data, yielding the maximum rate constant at 81 s Ϫ1 with K d,ATP at 3.3 M (Fig. 2B). This rate constant represents the rate of the ATPpromoted isomerization that occurs to generate the MT⅐KIF3*⅐ ATP intermediate poised for ATP hydrolysis (Scheme 1 and Table 1).
The amplitude (A 0 ) of the exponential phase represents the concentration of [␣-32 P]ATP bound at the active site that proceeded to form the product [␣-32 P]ADP⅐P i during the first ATP turnover. The observed amplitude from each transient was plotted as a function of increasing concentrations of ATP. The maximum amplitude at 0.8 M ADP⅐P i per M active site concentration indicates that approximately one ATP turnover occurred before KIF3AC detachment from MT (Fig. 2C).
To determine the time course of ATP hydrolysis (Fig. 2, D-F), the acid-quench experiments were performed at the same concentrations of the MT⅐KIF3AC complex as in the pulse-chase experiments, and 300 mM KCl (syringe concentration) was also added to the [␣-32 P]ATP syringe to slow steadystate ATP turnover to better define the first ATP turnover. The concentration of [␣-32 P]ADP⅐P i product formed was plotted as a function of time. Each transient was also biphasic with an initial exponential phase followed by a linear phase, indicative that a slow step occurred after ATP hydrolysis (Fig. 2D). The observed rates of the initial exponential phase were plotted as a function of increasing ATP concentration, and the hyperbolic fit to the data provided the maximum rate constant of ATP hydrolysis at 69 s Ϫ1 with the K d,ATP at 9.7 M (Fig. 2E and Table  1). Fig. 2F presents the burst amplitude data (A 0 ) of the exponential phase plotted as a function of ATP concentration, and the hyperbolic fit to the data provided A max of 0.77 M ADP⅐P i per M KIF3AC active site concentration. The amplitude data for both the pulse-chase kinetics and the acid-quench kinetics are consistent with the interpretation that each motor head (KIF3A or KIF3C) were able to bind and hydrolyze one ATP during the first ATP turnover.
In sum, these results from the pulse-chase and acid-quench experiments show that ATP binding and ATP hydrolysis for KIF3AC are both fast events. Furthermore, the observation that  the maximum amplitude at ϳ0.8 indicates that the experimental design were able to separate the first ATP turnover from the subsequent ATP turnovers despite the high processivity of KIF3AC.
KIF3AC and KIF3AA Microtubule Collision Is Significantly Faster than Microtubule Collision by KIF3CC-To measure the kinetics of MT collision and ADP release, we designed an experiment to measure native ADP release rather than mantADP release. Because the binding of mantATP is so rapid ( Fig. 1 and Table 1), we can use the fluorescence signal for mantATP binding as a readout of MT collision followed by ADP release. For this experiment, the KIF3 motors were rapidly mixed in the stopped-flow instrument with MTs plus mantATP (Fig. 3). At low MT concentrations, ADP release is limited by the collision of KIF3 with the MT; therefore, the linear fit of the data in this concentration range provides the second order rate constant for MT association. The rate constant of MT association by KIF3AC is 6.6 M Ϫ1 s Ϫ1 (Fig. 3B) with KIF3AA at 11.4 M Ϫ1 s Ϫ1 (Fig. 3C). In contrast, MT collision by KIF3CC is significantly slower at 2.1 M Ϫ1 s Ϫ1 (Fig. 3D). Note that the MT association rate of KIF3AC is intermediate between the constants for KIF3AA and KIF3CC (Fig. 3E).
The ADP Release Kinetics of KIF3AC Are Intermediate between KIF3AA and KIF3CC-As the MT concentration in these experiments is increased, the data become hyperbolic and represent the rate of ADP release that directly follows MT collision. The rate constant for KIF3AA was 78 s Ϫ1 , whereas KIF3CC exhibits a much slower maximum rate of ADP release at 8 s Ϫ1 (Fig. 3, C-E). In contrast, the corresponding rate constant of ADP release for KIF3AC was observed at 51 s Ϫ1 , which is intermediate between the constants for KIF3AA and KIF3CC (Fig. 3E).
The MT association and ADP release kinetics can be interpreted as follows. If there were an intrinsic bias of entry into the processive run (Fig. 5, E0 -E1), we would expect the kinetics of KIF3AC to be more similar to the kinetics of either KIF3AA or KIF3CC. In contrast, for both MT association and MT-promoted ADP release, the kinetics for both events for KIF3AC are intermediate between those for KIF3AA and KIF3CC, suggesting that either head, KIF3A or KIF3C, can begin the processive run. Note too that the transients for KIF3AC (Fig. 3A) are biphasic with the initial phase rapid followed by a slower phase at 4 -8 s Ϫ1 , and the amplitude is approximately split between the two phases. In contrast to KIF3AC, the transients for KIF3AA and KIF3CC are not strictly biphasic with the amplitude associated with the first exponential phase followed by a slow linear phase (Fig. 4).
Intermolecular Communication Modulates KIF3AC Entry into Processive Run-Our single molecule motility analysis (30) revealed that KIF3AC is more processive than KIF3AA or KIF3CC based on the run length data, implying that processivity is tuned in part by the intermolecular communication that occurs within the heterodimer. In addition, the results for MT association coupled with ADP release indicate a role for head-head communication because the behavior of KIF3AC is not equivalent to either KIF3AA or KIF3CC. To explore this hypothesis in greater detail, a direct comparison was made by measuring the kinetics of ADP release based on mantATP binding (Fig. 4). KIF3AC, KIF3AA, KIF3CC, and mixtures of KIF3AA ϩ KIF3CC at 5 M KIF3 site concentration were rapidly mixed with 30 M MTs ϩ 100 M mantATP (syringe concentrations). The final concentrations were 2.5 M KIF3 sites, 15 M MTs, and 50 M mantATP. The mixtures corresponded to 75% KIF3AA ϩ 25% KIF3CC, 50% KIF3AA ϩ 50% KIF3CC, and 25% KIF3AA ϩ 75% KIF3CC with the final motor concentration at 2.5 M after mixing (Fig. 4).
Note that for the KIF3AA ϩ KIF3CC mixture experiments, KIF3AA MT collision is independent of KIF3CC MT collision. The data reported in Fig. 4C revealed that 100% KIF3AA resulted in the fastest rate of ADP release at 68 s Ϫ1 , consistent show an exponential increase in fluorescence as a function of time, and a double exponential function fit to each provided the observed rates of each phase. B, the observed rates of the initial exponential phase of KIF3AC were plotted as a function of mantATP concentration, and the linear fit (Equation 1) to the data provided the second order rate constant for mantATP binding at 11.0 Ϯ 0.6 M Ϫ1 s Ϫ1 with k off ϭ 21.4 Ϯ 7.2 s Ϫ1 . The inset presents the observed rates of the second exponential phase, and the hyperbolic fit provides k max at 22.0 Ϯ 2.1 s Ϫ1 with K 1/2,mantATP at 8.6 Ϯ 1.9 M. C, the observed rates of the initial exponential phase were plotted as a function of mantATP concentration, and the linear fit to the data provided the second order rate constant for mantATP binding for KIF3AA at 16.0 Ϯ 0.5 M Ϫ1 s Ϫ1 with k off ϭ 10.5 Ϯ 5.1 s Ϫ1 and KIF3CC at 0.68 Ϯ 0.04 M Ϫ1 s Ϫ1 with k off ϭ 7.7 Ϯ 0.4 s Ϫ1 in comparison with the data for KIF3AC also shown in B. FEBRUARY 26, 2016 • VOLUME 291 • NUMBER 9

Initiation of the Processive Run by KIF3AC
with the results in Fig. 3 at 15 M MTs. Furthermore, each mixture of KIF3AA ϩ KIF3CC resulted in an initial fast phase more similar to the rate for KIF3AA than KIF3AC at 36 s Ϫ1 and a slow phase of 4 -5 s Ϫ1 . In fact, the 50% KIF3AA ϩ 50% KIF3CC mixture resulted in a rate for ADP release at 62 s Ϫ1 for the initial fast phase attributed to KIF3AA and a second slow phase at 4.8 s Ϫ1 attributed to KIF3CC. Therefore, these results suggest that the kinetics for KIF3AC were not the sum of the kinetics of the independent KIF3A and KIF3C motor heads.
If one looks directly at the transients in Fig. 4 (A and B), there are differences in the amplitude associated with the initial fast and slow phases. For KIF3AA, almost all the amplitude is associated with the initial fast phase, indicating that ADP release from the first head was not well separated from ADP release from the second head. In contrast, for KIF3AC, the amplitude is split between the two phases, implying that the initial fast phase represents KIF3A and the slow phase represents KIF3C. The transients from the mixtures do show an initial fast phase that is similar to KIF3AA, but there is a progressive decrease in the total amplitude as the percentage of KIF3CC increases. Previous experiments for Fig. 3D showed that for KIF3CC, the slow rate of ADP release required a 1-s time domain for the stoppedflow experiment. Therefore, we attribute the decrease in amplitude for the mixtures and KIF3CC to be due to the shorter 0.5-s time domain of this experiment where the full amplitude cannot be observed. Importantly, the mixture of 50% KIF3AA plus 50% KIF3CC did not recapitulate the kinetics of 100% KIF3AC either in direct comparison of the transients or in comparison of the rates of ADP release (Fig. 4C). If the experimental transient for 50% KIF3AA ϩ 50% KIF3CC accurately represents the contribution from each motor head independently colliding with the MT, then a similar transient can be generated computationally using the data from the transients of 100% KIF3AA and 100% KIF3CC. Each transient represents 1,000 data points, and at each time point for the mixtures, it is the sum of the contribution from KIF3A plus KIF3C. Fig. 4B shows both the computationally generated transient for 50% KIF3AA ϩ 50% KIF3CC (gold) relative to the experimental transient in green. Note that they superimpose on each other. Moreover, both the experimental and the computationally derived transients do not recapitulate the transient for KIF3AC (red).
This analysis supports our conclusion that MT collision by KIF3AA and KIF3CC in the mixture experiments represents independent events, and intermolecular communication between KIF3A and KIF3C can only occur within the heterodimeric KIF3AC. Furthermore, the results from Figs. 3 and 4 also reinforce the hypothesis that although either KIF3A or KIF3C can initiate the processive run, the influence of KIF3A on KIF3C and KIF3C on KIF3A is due to the intermolecular communication.

Discussion
The stepping model proposed in Fig. 5 illustrates the sequence of events by which a processive kinesin couples ATP turnover to each step along the MT. The cycle begins as one motor head of the dimer collides with the MT followed by ADP release (E0 -E1). The asymmetry in the dimer is now established by which ATP turnover on one head is maintained out of phase with ATP turnover on the partner head. At E1, ATP binding at the leading head generates a series of structural transitions including neck linker docking onto the catalytic core and movement of the trailing ADP head forward to its next MT binding site, 16 nm ahead toward the MT plus end (E1-E4). ADP is subsequently released, which tightens the affinity of the leading head for the MT, thereby generating strain because the neck linkers are now in opposite orientations, and both heads are tightly bound to the MT (E4 -E5). ATP hydrolysis is triggered on the trailing head followed by phosphate release coupled with detachment of the trailing head from the MT (E6 -E7). This model proposes that ATP binding at the leading head of the E5 intermediate is inhibited because of the backward orientation of its neck linker and the intermolecular tension generated by the helix ␣7 coiled-coil (30, 34 -38). The second step begins as the partner head binds ATP at E7.
For the homodimeric processive kinesins, the two motor domains in solution each have ADP tightly bound, and the asymmetry in the ATPase cycle is not established until MT collision. Therefore, a key question for heterodimeric KIF3AC is whether there is an intrinsic asymmetry in the heterodimer when in solution that favors KIF3A or KIF3C to initiate the processive run. Entry into the Processive Run-The results presented indicate that either KIF3A or KIF3C can begin the processive run. The results in Fig. 3  However, at first glance the pulse-chase and acid quench kinetics for ATP binding and ATP hydrolysis appear to suggest that it is not so simple where the rates at each step are the averages of KIF3AA and KIF3CC. The results from Fig. 2 show that for KIF3AC the ATP-promoted isomerization at 81 s Ϫ1 to form the intermediate poised for ATP hydrolysis is a fast step   (E0 and E1). The leading head is nucleotide-free with the trailing head detached from the MT and ADP tightly bound. ATP binding at the leading head triggers a series of structural transitions including neck linker docking that promotes the trailing head to move forward to its next MT binding site (E2-E4). ADP is released upon MT collision, resulting in the E5 two-head-bound state. Strain develops between the two heads and inhibits ATP binding to the nucleotide-free leading head (E5). ATP hydrolysis on the trailing head followed by phosphate (P i ) release relieves the interhead tension (E5-E7). The trailing head with ADP tightly bound has a weak affinity for the MT and thereby detaches (E7). The leading head can now bind ATP, which initiates the next step coupled to ATP turnover without detachment from the MT. followed by ATP hydrolysis at 69 s Ϫ1 . Both of these rate constants are very fast. Furthermore, the maximum amplitude at ϳ0.8 per active site in each experiment suggests that both KIF3A and KIF3C were able to bind and hydrolyze ATP rapidly. If these rate constants only represented the contribution by KIF3A with exceedingly slow ATP binding and hydrolysis by KIF3C, then the amplitude expected would have been ϳ0.5 per site representing only the KIF3A head. Note too that the KIF3CC second order rate constant for mantATP binding (Fig.  1) at 0.68 M Ϫ1 s Ϫ1 predicts an ATP binding rate at 100 M ATP of ϳ64 s Ϫ1 . Therefore, these data indicate that the kinetics observed for KIF3AC are consistent with the hypothesis that either KIF3A or KIF3C can initiate the processive run.
Impact of KIF3A on KIF3C and KIF3C on KIF3A-If the KIF3AC kinetics of MT collision and ADP release were simply the average of both heads as suggested for KIF3AB (31), then the kinetics for these steps should be consistent with the kinetics of a 1:1 mixture of homodimeric KIF3AA ϩ KIF3CC equivalent in concentration to KIF3AC. The data in Fig. 4 clearly show that we were unable to generate a KIF3AC-like transient by using mixtures of KIF3AA ϩ KIF3CC. Although the mixture of 25% KIF3AA ϩ 75% KIF3CC (Fig. 4A, blue transient) was similar to the initial part of the KIF3AC transient (red), it failed to achieve the full amplitude of the initial exponential phase or a second exponential phase similar to that of KIF3AC. Comparison of the KIF3AC transient to the transient from 50% KIF3AA ϩ 50% KIF3CC (green) also shows that the mixture of the homodimers did not recapitulate the kinetics of KIF3AC. Therefore, these results show that although the processive run may begin with either KIF3A or KIF3C motor domain collision with the MT, the kinetics observed are an emergent property caused by intermolecular communication within the heterodimer rather than the average of the intrinsic behavior of each motor head.
Key Questions Ahead-Although these initial studies have provided a much better understanding of entry into the processive run, there is much that is still unknown. One key question is whether once in a processive run, the KIF3A and KIF3C heads retain their intrinsic mechanochemistry or rather because of the heterodimerization, do KIF3A and KIF3C become similar with a unique emergent mechanochemistry because of intermolecular communication. The results presented here, the single molecule velocity rates, and the steady-state ATPase k cat parameters suggest that intermolecular interactions within the heterodimer result in each motor head regulating the other (Table 1 and Ref. 30). Thus, we may discover that the influence of KIF3A on KIF3C and KIF3C on KIF3A may yield a heterodimer that behaves not so differently from a homodimer after all.
Author Contributions-P. Z., I. R., and S. P. G. designed the research; P. Z. performed the research; and P. Z., I. R., and S. P. G. analyzed the data and wrote the manuscript.